Conventionally, moving image encoding such as moving picture experts group phase1 (MPEG1), MPEG2, H.264 or H.265 is configured to determine which of a motion compensating prediction error signal and an in-frame pixel signal it is advantageous to subject to discrete cosine transform (DCT) for each small block (16×16). Determination is made using one having smaller in-block variance. A selected signal is subjected to DCT conversion, for example. The DCT coefficient is quantized in a quantization step set for each block, and then allocates a variable length code to a quantization representative value. For allocating a variable length code also to motion vector information, a code is allocated to a difference vector from a vector in the vicinity, for improvement in efficiency.
As a citation list, PATENT LITERATURE 1 discloses a technique of a compressed image processing device which has an image recognition function or an image composition function, for example, for data of a compressed and coded image and can exchange background image data.